Modelling Smooth Transition (Auto)Regression in Rental Prices of U.K. Real Estate Sectors

Abstract

This paper presents a new approach to analysing real estate rental adjustment processes. We model the rental adjustment process in U.K. commercial real estate sectors by using non-linear smooth transition (auto)regression (ST(A)R) models. The ST(A)R models, which clearly outperform their linear counterparts, account for different phases or regimes, which can be considered subperiods of the physical real estate market cycle. Because the physical market cycles are the result of supply and demand interactions reflected in substantial changes in economic factors, it is of great importance to investors to understand the drivers of these cyclical shifts. The results are striking: In addition to the influence of contemporaneous exogenous factors, rental growth is driven by both its own past and by lagged exogenous factors, exhibiting the well-known stickiness of real estate markets. Furthermore, the rental series behave differently in different phases of the market cycle, and the adjustment of real rents is best described in a non-linear fashion. Empirical evidence shows that models using lagged real rental growth rates as the transition variable seem to be adequate for explaining and forecasting the rental adjustments process. A transition into another state sets in for the industrial, retail, and office sectors at real rental growth rates above 2.5%, 4.2%, and 6.1% respectively. Using the ST(A)R models enables us to describe the very nature of the rental adjustment process more accurately, with the technical properties of the estimation method being more in line with theoretical underpinnings than linear methods, without state-dependent or time-varying features.